Project: Investigate a Dataset

Investigation of the Soccer Dataset

Table of Contents

• Introduction
• Data Wrangling
• Exploratory Data Analysis
• Conclusions

Introduction

This investigation uses the football dataset provided through this link: (https://www.kaggle.com/datasets/hugomathien/soccer). The perspective taken was that of analysing the creative attributes of the teams that win championships. The English Premier League was selected as the focus with the period of 2009 to 2015.

The pandas documentation was utilized for reference (https://pandas.pydata.org/docs/)

The plotly documentation was utilized for plotting reference (https://plotly.com/python/radar-chart/)

Research Question 1: Are premier league winning teams characterized primarily by creativity, team cohesion and fluidity?

Research Question 2: Based on the result of question 1, can we then claim that seasons where the champions were not at the top of the creativity charts as being more competitive?

# Use this cell to set up import statements for all of the packages that you
#   plan to use.

# Remember to include a 'magic word' so that your visualizations are plotted
#   inline with the notebook. See this page for more:
#   http://ipython.readthedocs.io/en/stable/interactive/magics.html

import pandas as pd
# import sqlite3
import numpy as np
import matplotlib.pyplot as plt
import plotly.express as px
import plotly.graph_objects as go
%matplotlib inline

Data Wrangling

Tip: In this section of the report, you will load in the data, check for cleanliness, and then trim and clean your dataset for analysis. Make sure that you document your steps carefully and justify your cleaning decisions.

General Properties

# Loading the relevant files
df_T = pd.read_csv('Team.csv')
df_TA = pd.read_csv('Team_Attributes.csv')
  
# Creating a dataframe join
df = pd.merge(df_T, df_TA, on='team_api_id', how='left')

# Verify that result of SQL query is stored in the dataframe
df.head()

	id_x	team_api_id	team_fifa_api_id_x	team_long_name	team_short_name	id_y	team_fifa_api_id_y	date	buildUpPlaySpeed	buildUpPlaySpeedClass	...	chanceCreationShooting	chanceCreationShootingClass	chanceCreationPositioningClass	defencePressure	defencePressureClass	defenceAggression	defenceAggressionClass	defenceTeamWidth	defenceTeamWidthClass	defenceDefenderLineClass
0	1	9987	673.0	KRC Genk	GEN	485.0	673.0	2010-02-22 00:00:00	45.0	Balanced	...	60.0	Normal	Organised	70.0	High	65.0	Press	70.0	Wide	Cover
1	1	9987	673.0	KRC Genk	GEN	486.0	673.0	2011-02-22 00:00:00	66.0	Balanced	...	51.0	Normal	Organised	48.0	Medium	47.0	Press	54.0	Normal	Offside Trap
2	1	9987	673.0	KRC Genk	GEN	487.0	673.0	2012-02-22 00:00:00	53.0	Balanced	...	56.0	Normal	Organised	47.0	Medium	45.0	Press	55.0	Normal	Cover
3	1	9987	673.0	KRC Genk	GEN	488.0	673.0	2013-09-20 00:00:00	58.0	Balanced	...	56.0	Normal	Organised	47.0	Medium	45.0	Press	55.0	Normal	Cover
4	1	9987	673.0	KRC Genk	GEN	489.0	673.0	2014-09-19 00:00:00	58.0	Balanced	...	56.0	Normal	Organised	47.0	Medium	45.0	Press	55.0	Normal	Cover

5 rows × 29 columns

# Shape of the data
df.shape

(1469, 29)

# Description of the data
df.describe()

	id_x	team_api_id	team_fifa_api_id_x	id_y	team_fifa_api_id_y	buildUpPlaySpeed	buildUpPlayDribbling	buildUpPlayPassing	chanceCreationPassing	chanceCreationCrossing	chanceCreationShooting	defencePressure	defenceAggression	defenceTeamWidth
count	1469.000000	1469.000000	1458.000000	1458.000000	1458.000000	1458.000000	489.000000	1458.000000	1458.000000	1458.000000	1458.000000	1458.000000	1458.000000	1458.000000
mean	22750.390061	10213.878829	17706.982167	729.500000	17706.982167	52.462277	48.607362	48.490398	52.165295	53.731824	53.969136	46.017147	49.251029	52.185871
std	15032.343553	14675.457811	39179.857739	421.032659	39179.857739	11.545869	9.678290	10.896101	10.360793	11.086796	10.327566	10.227225	9.738028	9.574712
min	1.000000	1601.000000	1.000000	1.000000	1.000000	20.000000	24.000000	20.000000	21.000000	20.000000	22.000000	23.000000	24.000000	29.000000
25%	9548.000000	8456.000000	110.000000	365.250000	110.000000	45.000000	42.000000	40.000000	46.000000	47.000000	48.000000	39.000000	44.000000	47.000000
50%	20525.000000	8674.000000	485.000000	729.500000	485.000000	52.000000	49.000000	50.000000	52.000000	53.000000	53.000000	45.000000	48.000000	52.000000
75%	35295.000000	9904.000000	1900.000000	1093.750000	1900.000000	62.000000	55.000000	55.000000	59.000000	62.000000	61.000000	51.000000	55.000000	58.000000
max	51606.000000	274581.000000	112513.000000	1458.000000	112513.000000	80.000000	77.000000	80.000000	80.000000	80.000000	80.000000	72.000000	72.000000	73.000000

# View the index number and label for each column
def view_index():
    for i, v in enumerate(df.columns):
        print(i, v)

# View the index number and label for each column
view_index()

0 id_x
1 team_api_id
2 team_fifa_api_id_x
3 team_long_name
4 team_short_name
5 id_y
6 team_fifa_api_id_y
7 date
8 buildUpPlaySpeed
9 buildUpPlaySpeedClass
10 buildUpPlayDribbling
11 buildUpPlayDribblingClass
12 buildUpPlayPassing
13 buildUpPlayPassingClass
14 buildUpPlayPositioningClass
15 chanceCreationPassing
16 chanceCreationPassingClass
17 chanceCreationCrossing
18 chanceCreationCrossingClass
19 chanceCreationShooting
20 chanceCreationShootingClass
21 chanceCreationPositioningClass
22 defencePressure
23 defencePressureClass
24 defenceAggression
25 defenceAggressionClass
26 defenceTeamWidth
27 defenceTeamWidthClass
28 defenceDefenderLineClass

Data Cleaning

Check for duplicates within the dataset

# Check for duplicates
sum(df.duplicated())

0

Check for columns with null values

# Check for null values
df.isnull().sum()

id_x                                0
team_api_id                         0
team_fifa_api_id_x                 11
team_long_name                      0
team_short_name                     0
id_y                               11
team_fifa_api_id_y                 11
date                               11
buildUpPlaySpeed                   11
buildUpPlaySpeedClass              11
buildUpPlayDribbling              980
buildUpPlayDribblingClass          11
buildUpPlayPassing                 11
buildUpPlayPassingClass            11
buildUpPlayPositioningClass        11
chanceCreationPassing              11
chanceCreationPassingClass         11
chanceCreationCrossing             11
chanceCreationCrossingClass        11
chanceCreationShooting             11
chanceCreationShootingClass        11
chanceCreationPositioningClass     11
defencePressure                    11
defencePressureClass               11
defenceAggression                  11
defenceAggressionClass             11
defenceTeamWidth                   11
defenceTeamWidthClass              11
defenceDefenderLineClass           11
dtype: int64

Replace the columns with null values

Since the dataset is primarily numerict, we can replace the null values with zero for a more seamless set of data types

We will only replace the data for the columns we picked to answer our questions. Later we will drop the columns that aren't used

# Replace null values with zero per relevant row
df["team_fifa_api_id_x"].fillna(0, inplace = True)
df["date"].fillna(0, inplace = True)
df["buildUpPlaySpeed"].fillna(0, inplace = True)
df["chanceCreationPassing"].fillna(0, inplace = True)
df["chanceCreationCrossing"].fillna(0, inplace = True)
df["chanceCreationShooting"].fillna(0, inplace = True)

Confirm the replacement worked on the relevant columns

# Check if null values were replaced
df.isnull().sum()

id_x                                0
team_api_id                         0
team_fifa_api_id_x                  0
team_long_name                      0
team_short_name                     0
id_y                               11
team_fifa_api_id_y                 11
date                                0
buildUpPlaySpeed                    0
buildUpPlaySpeedClass              11
buildUpPlayDribbling              980
buildUpPlayDribblingClass          11
buildUpPlayPassing                 11
buildUpPlayPassingClass            11
buildUpPlayPositioningClass        11
chanceCreationPassing               0
chanceCreationPassingClass         11
chanceCreationCrossing              0
chanceCreationCrossingClass        11
chanceCreationShooting              0
chanceCreationShootingClass        11
chanceCreationPositioningClass     11
defencePressure                    11
defencePressureClass               11
defenceAggression                  11
defenceAggressionClass             11
defenceTeamWidth                   11
defenceTeamWidthClass              11
defenceDefenderLineClass           11
dtype: int64

Finally we drop the unnecessary columns from our dataset to better focus on our relevant data

# Dropping unnecessary columns
dropped_columns = ['id_x', 'id_y', 'team_short_name', 'team_fifa_api_id_y', 'buildUpPlaySpeedClass', 'buildUpPlayDribbling', 
           'buildUpPlayDribblingClass', 'buildUpPlayPassing', 'buildUpPlayPassingClass', 'buildUpPlayPositioningClass', 
          'chanceCreationPassingClass', 'chanceCreationShootingClass', 'chanceCreationCrossingClass', 'chanceCreationShootingClass', 
           'chanceCreationPositioningClass', 'defencePressure', 'defencePressureClass', 'defenceAggression', 'defenceAggressionClass', 
           'defenceTeamWidth', 'defenceTeamWidthClass', 'defenceDefenderLineClass']
df.drop(dropped_columns, inplace=True, axis=1)

Confirmation of the columns present in the new dataset

# View the index number and label for each column
view_index()

0 team_api_id
1 team_fifa_api_id_x
2 team_long_name
3 date
4 buildUpPlaySpeed
5 chanceCreationPassing
6 chanceCreationCrossing
7 chanceCreationShooting

# creation of the attributes dataframe
df_A = df
df_A.head()

	team_api_id	team_fifa_api_id_x	team_long_name	date	buildUpPlaySpeed	chanceCreationPassing	chanceCreationCrossing	chanceCreationShooting
0	9987	673.0	KRC Genk	2010-02-22 00:00:00	45.0	50.0	35.0	60.0
1	9987	673.0	KRC Genk	2011-02-22 00:00:00	66.0	65.0	66.0	51.0
2	9987	673.0	KRC Genk	2012-02-22 00:00:00	53.0	55.0	48.0	56.0
3	9987	673.0	KRC Genk	2013-09-20 00:00:00	58.0	67.0	48.0	56.0
4	9987	673.0	KRC Genk	2014-09-19 00:00:00	58.0	67.0	48.0	56.0

Save the new dataset to a csv file that we will use from this point forward

# save this for later
df_A.to_csv('football_dataset_creativity.csv', index=False)

df_fdc = pd.read_csv('football_dataset_creativity.csv')
df_fdc.head()

	team_api_id	team_fifa_api_id_x	team_long_name	date	buildUpPlaySpeed	chanceCreationPassing	chanceCreationCrossing	chanceCreationShooting
0	9987	673.0	KRC Genk	2010-02-22 00:00:00	45.0	50.0	35.0	60.0
1	9987	673.0	KRC Genk	2011-02-22 00:00:00	66.0	65.0	66.0	51.0
2	9987	673.0	KRC Genk	2012-02-22 00:00:00	53.0	55.0	48.0	56.0
3	9987	673.0	KRC Genk	2013-09-20 00:00:00	58.0	67.0	48.0	56.0
4	9987	673.0	KRC Genk	2014-09-19 00:00:00	58.0	67.0	48.0	56.0

# Description of the data
df_fdc.describe()

	team_api_id	team_fifa_api_id_x	buildUpPlaySpeed	chanceCreationPassing	chanceCreationCrossing	chanceCreationShooting
count	1469.000000	1469.000000	1469.000000	1469.000000	1469.000000	1469.000000
mean	10213.878829	17574.390742	52.069435	51.774677	53.329476	53.565010
std	14675.457811	39062.649049	12.360307	11.259644	11.977799	11.292526
min	1601.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	8456.000000	109.000000	45.000000	46.000000	47.000000	48.000000
50%	8674.000000	483.000000	52.000000	52.000000	53.000000	53.000000
75%	9904.000000	1897.000000	62.000000	59.000000	62.000000	61.000000
max	274581.000000	112513.000000	80.000000	80.000000	80.000000	80.000000

# Info of the data
df_fdc.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1469 entries, 0 to 1468
Data columns (total 8 columns):
 #   Column                  Non-Null Count  Dtype  
---  ------                  --------------  -----  
 0   team_api_id             1469 non-null   int64  
 1   team_fifa_api_id_x      1469 non-null   float64
 2   team_long_name          1469 non-null   object 
 3   date                    1469 non-null   object 
 4   buildUpPlaySpeed        1469 non-null   float64
 5   chanceCreationPassing   1469 non-null   float64
 6   chanceCreationCrossing  1469 non-null   float64
 7   chanceCreationShooting  1469 non-null   float64
dtypes: float64(5), int64(1), object(2)
memory usage: 91.9+ KB

Rename the columns

team_long_name refers to the football clubs full name

team_fifa_api_id_x refers to the identifier given by FIFA to the football club for reference

team_api_id refers to the identifier for each football club for reference

date this is the record of the date i.e. the season when the data was recorded

buildUpPlaySpeed refers to the speed/rate a team takes link up play leading to the creation of a goal scoring opportunity (this figure would most likely be an average of all the matches within a season)

chanceCreationPassing refers to the creation of passing chances per season (this figure would most likely be an average of all the matches within a season)

chanceCreationCrossing refers to the creation of crossing chances per season (this figure would most likely be an average of all the matches within a season)

chanceCreationShooting refers to the creation of shooting chances per season (this figure would most likely be an average of all the matches within a season)

# Operations on the data
# Rename the columns
df_fdc.rename(columns = {'team_long_name':'team_name', 'team_fifa_api_id_x':'fifa_id', 'team_api_id':'team_id'}, inplace = True)
# Truncate the date column for the plot
df_fdc['date'] = df_fdc['date'].str[:4]
# View the index number and label for each column
view_index()

0 team_api_id
1 team_fifa_api_id_x
2 team_long_name
3 date
4 buildUpPlaySpeed
5 chanceCreationPassing
6 chanceCreationCrossing
7 chanceCreationShooting

Exploratory Data Analysis

Research Question 1: Are premier league winning teams characterized primarily by creativity, team cohesion and fluidity?

A comparitive analysis on the EPL Champions between 2009 and 2015

Team Attributes being compared:

1. Build up play speed
2. Chance creation (passing)
3. Chance creation (crossing)
4. Chance creation (shooting)

df_Chelsea = df_fdc[df_fdc["team_name"] == "Chelsea"]
df_M_United = df_fdc[df_fdc["team_name"] == "Manchester United"]
df_M_City = df_fdc[df_fdc["team_name"] == "Manchester City"]

Line graph plots for the data

# Line plot function

def line_plot(p1,p2,p3,title,xlabel,ylabel,label_10,label_11,label_12,label_13,label_14,label_15):
    x1 = df_Chelsea['date']
    x2 = df_M_United['date']
    x3 = df_M_City['date']
    
    y1 = p1
    y2 = p2
    y3 = p3
    
    plt.title(title)
    plt.xlabel(xlabel)
    plt.ylabel(ylabel)
    
    Chelsea, = plt.plot(x1, y1, label="Chelsea", marker='o', color='blue')
    M_United, = plt.plot(x2, y2, label="Manchester United", marker='o', color='red')
    M_City, = plt.plot(x3, y3, label="Manchester City", marker='o', color='green')
    
    #Champion Label
    label_10
    label_11
    label_12
    label_13
    label_14
    label_15
    
    leg = plt.legend(loc='lower left')
    plt.show()

# plot distribution of build up play speed
line_plot(df_Chelsea['buildUpPlaySpeed'], 
          df_M_United['buildUpPlaySpeed'], 
          df_M_City['buildUpPlaySpeed'], 
          'EPL Champions Build Up Play Between 2010 and 2015', 
          'Year', 
          'Rate of Build Up Play (Speed)', 
          plt.text(0,70,'Chelsea'), 
          plt.text(1,65,'United'), 
          plt.text(2,45,'City'), 
          plt.text(3,46,'United'), 
          plt.text(4,59,'City'), 
          plt.text(5,67,'Chelsea'))

Based on the results from the chart of build up play speed, we may conclude that our initial question does not always fit as in the 2011, 2012 and 2014, the EPL champions were not as direct in their build up play compared to the other two teams that were frequent winners within the period of 2010 to 2015

# plot distribution of passing chance creation
line_plot(df_Chelsea['chanceCreationPassing'], 
          df_M_United['chanceCreationPassing'], 
          df_M_City['chanceCreationPassing'], 
          'EPL Champions Pass Creation Between 2010 and 2015', 
          'Year', 
          '# Chances Created (Passing)', 
          plt.text(0,56,'Chelsea'),
          plt.text(1,65,'United'),
          plt.text(2,40,'City'),
          plt.text(3,46,'United'),
          plt.text(4,32,'City'),
          plt.text(5,41,'Chelsea'))

For passing chance creation the teams that won the EPL in the 2011, 2012, 2014 and 2015 season weren't at the top of the passing chance creation charts. The seasons 2011, 2012 and 2014 are present once more.

# plot distribution of crossing chance creation
line_plot(df_Chelsea['chanceCreationCrossing'], 
          df_M_United['chanceCreationCrossing'], 
          df_M_City['chanceCreationCrossing'], 
          'EPL Champions Cross Creation Between 2010 and 2015', 
          'Year', 
          '# Chances Created (Crossing)', 
          plt.text(0,70,'Chelsea'),
          plt.text(1,65,'United'),
          plt.text(2,38,'City'),
          plt.text(3,68,'United'),
          plt.text(4,49,'City'),
          plt.text(5,34,'Chelsea'))

Cross chance creation starts of at a high rate before dropping off for all teams as the seasons progress. In this case the seasons are 2012 and 2014 common in the previous visualizations where the champion doesn't lead the charts.

# plot distribution of shooting chance creation
line_plot(df_Chelsea['chanceCreationShooting'], 
          df_M_United['chanceCreationShooting'], 
          df_M_City['chanceCreationShooting'], 
          'EPL Champions Shot Creation Between 2010 and 2015', 
          'Year', 
          '# Chances Created (Shots)', 
          plt.text(0,70,'Chelsea'),
          plt.text(1,70,'United'),
          plt.text(2,48,'City'),
          plt.text(3,37,'United'),
          plt.text(4,62,'City'),
          plt.text(5,44,'Chelsea'))

For the seasons 2011, 2012, 2013 and 2014 the champion does not produce the most shot creation chances.

Histogram Plots for the data

# Bar plot function
def bar_plot(x1,x2,x3,title,ylabel,label_10,label_11,label_12,label_13,label_14,label_15):
    N = 6
    ind = np.arange(N) 
    width = 0.25

    Chelsea = x1
    M_United = x2
    M_City = x3
    
    bar1 = plt.bar(ind, Chelsea, width, color = 'b')
    bar2 = plt.bar(ind+width, M_United, width, color='r')
    bar3 = plt.bar(ind+width*2, M_City, width, color = 'g')

    plt.title(title)
    plt.xlabel("Year")
    plt.ylabel(ylabel)
    
    #Champion Label
    label_10
    label_11
    label_12
    label_13
    label_14
    label_15

    plt.xticks(ind+width,['2010', '2011', '2012', '2013', '2014', '2015'])
    plt.legend((bar1, bar2, bar3), ('Chelsea', 'United', 'City'))
    plt.show()

# plot a bar graph for build up play speed
bar_plot(df_Chelsea['buildUpPlaySpeed'], 
          df_M_United['buildUpPlaySpeed'], 
          df_M_City['buildUpPlaySpeed'], 
          'EPL Champions Build Up Play Between 2010 and 2015',
          'Rate of Build Up Play (Speed)',
         plt.text(0,70,'Chelsea'),
         plt.text(1,65,'United'), 
         plt.text(2,45,'City'), 
         plt.text(3,46,'United'), 
         plt.text(4,59,'City'), 
         plt.text(5,67,'Chelsea'))

Based on the results from the chart of build up play speed, we may conclude that our initial question does not always fit as in the 2011, 2012 and 2014, the EPL champions were not as direct in their build up play compared to the other two teams that were frequent winners within the period of 2010 to 2015

# plot a bar graph for chance creation passing
bar_plot(df_Chelsea['chanceCreationPassing'], 
          df_M_United['chanceCreationPassing'], 
          df_M_City['chanceCreationPassing'], 
          'EPL Champions Chance Created (Passes) Between 2010 and 2015',
          '# Passing Chances Created',
          plt.text(0,56,'Chelsea'),
          plt.text(1,65,'United'),
          plt.text(2,40,'City'),
          plt.text(3,46,'United'),
          plt.text(4,32,'City'),
          plt.text(5,41,'Chelsea'))

For passing chance creation the teams that won the EPL in the 2011, 2012, 2014 and 2015 season weren't at the top of the passing chance creation charts. The seasons 2011, 2012 and 2014 are present once more.

# plot a bar graph for chance creation crossing
bar_plot(df_Chelsea['chanceCreationCrossing'], 
          df_M_United['chanceCreationCrossing'], 
          df_M_City['chanceCreationCrossing'], 
          'EPL Champions Chance Created (Crosses) Between 2010 and 2015',
          '# Crossing Chances Created', 
          plt.text(0,70,'Chelsea'),
          plt.text(1,65,'United'),
          plt.text(2,38,'City'),
          plt.text(3,68,'United'),
          plt.text(4,49,'City'),
          plt.text(5,34,'Chelsea'))

Cross chance creation starts of at a high rate before dropping off for all teams as the seasons progress. In this case the seasons are 2012 and 2014 common in the previous visualizations where the champion doesn't lead the charts.

# plot a bar graph for chance creation shooting
bar_plot(df_Chelsea['chanceCreationShooting'], 
          df_M_United['chanceCreationShooting'], 
          df_M_City['chanceCreationShooting'], 
          'EPL Champions Chance Created (Shots) Between 2010 and 2015',
          '# Shooting Chances Created', 
          plt.text(0,70,'Chelsea'),
          plt.text(1,70,'United'),
          plt.text(2,48,'City'),
          plt.text(3,37,'United'),
          plt.text(4,62,'City'),
          plt.text(5,44,'Chelsea'))

For the seasons 2011, 2012, 2013 and 2014 the champion does not produce the most shot creation chances.

# Combined Radar Charts for build up play speed
years = ['2010','2011','2012','2013','2014','2015']

fig = go.Figure()

fig.add_trace(go.Scatterpolar(
      r=df_Chelsea['buildUpPlaySpeed'],
      theta=years,
      fill='toself',
      name='Chelsea'
))
fig.add_trace(go.Scatterpolar(
      r=df_M_United['buildUpPlaySpeed'],
      theta=years,
      fill='toself',
      name='Manchester United'
))
fig.add_trace(go.Scatterpolar(
      r=df_M_City['buildUpPlaySpeed'],
      theta=years,
      fill='toself',
      name='Manchester City'
))
fig.update_layout(
  polar=dict(
    radialaxis=dict(
      visible=True,
      range=[0, 100]
    )),
  showlegend=False
)
fig.show()

The radar chart above represents the progression over the period of 2010 to 2015 for build up play speed per club. Hovering above the coloured sections will relay the club.

# Combined Radar Charts for chance creation passing
years = ['2010','2011','2012','2013','2014','2015']

fig = go.Figure()

fig.add_trace(go.Scatterpolar(
      r=df_Chelsea['chanceCreationPassing'],
      theta=years,
      fill='toself',
      name='Chelsea'
))
fig.add_trace(go.Scatterpolar(
      r=df_M_United['chanceCreationPassing'],
      theta=years,
      fill='toself',
      name='Manchester United'
))
fig.add_trace(go.Scatterpolar(
      r=df_M_City['chanceCreationPassing'],
      theta=years,
      fill='toself',
      name='Manchester City'
))
fig.update_layout(
  polar=dict(
    radialaxis=dict(
      visible=True,
      range=[0, 100]
    )),
  showlegend=False
)
fig.show()

The radar chart above represents the progression over the period of 2010 to 2015 for chance creation for passing per club. Hovering above the coloured sections will relay the club.

# Combined Radar Charts for chance creation crossing
years = ['2010','2011','2012','2013','2014','2015']

fig = go.Figure()

fig.add_trace(go.Scatterpolar(
      r=df_Chelsea['chanceCreationCrossing'],
      theta=years,
      fill='toself',
      name='Chelsea'
))
fig.add_trace(go.Scatterpolar(
      r=df_M_United['chanceCreationCrossing'],
      theta=years,
      fill='toself',
      name='Manchester United'
))
fig.add_trace(go.Scatterpolar(
      r=df_M_City['chanceCreationCrossing'],
      theta=years,
      fill='toself',
      name='Manchester City'
))
fig.update_layout(
  polar=dict(
    radialaxis=dict(
      visible=True,
      range=[0, 100]
    )),
  showlegend=False
)
fig.show()

The radar chart above represents the progression over the period of 2010 to 2015 for chance creation for crossing per club. Hovering above the coloured sections will relay the club.

# Combined Radar Charts for chance creation shooting
years = ['2010','2011','2012','2013','2014','2015']

fig = go.Figure()

fig.add_trace(go.Scatterpolar(
      r=df_Chelsea['chanceCreationShooting'],
      theta=years,
      fill='toself',
      name='Chelsea'
))
fig.add_trace(go.Scatterpolar(
      r=df_M_United['chanceCreationShooting'],
      theta=years,
      fill='toself',
      name='Manchester United'
))
fig.add_trace(go.Scatterpolar(
      r=df_M_City['chanceCreationShooting'],
      theta=years,
      fill='toself',
      name='Manchester City'
))
fig.update_layout(
  polar=dict(
    radialaxis=dict(
      visible=True,
      range=[0, 100]
    )),
  showlegend=False
)
fig.show()

The radar chart above represents the progression over the period of 2010 to 2015 for chance creation for shooting per club. Hovering above the coloured sections will relay the club.

Conclusions

Within all the visualizations the seasons most present are: 2012 - 4 occurrences 2014 - 4 occurrences

In the 2012 season the Champion Manchester City only won by goal difference while in the 2014 season the Champion Manchester City once again won narrowly by a 2 point gap. Thus in most cases between 2010 and 2015, the EPL champion is not the leader in either of the team attributes studied, with the seasons 2012 and 2014 showing greater competition towards the winning of the title.

The limitation to this exploration is that an exact correlation to a teams creative dominance to being a champion isn't a clear, thus leading this exploration to be more speculatory. Since the exploration is focussed on the eventual champions teams that are known to have creative football philosophies aren't represented within this study.